In re Meta Platforms, Inc., Securities Litigation

Opinion

OHIO PUBLIC EMPLOYEES Case No. 21-cv-08812-AMO RETIREMENT SYSTEM, Plaintiff, ORDER GRANTING IN PART v. MOTION TO DISMISS META PLATFORMS, INC., et al., Re: Dkt. No. 110 Defendants.

This securities case is about Meta and its senior executives’ alleged misstatements about harms that flow from its various social media platforms. Defendant’s motion to dismiss was heard before this Court on August 17, 2023. Having read the papers filed by the parties and carefully considered their arguments therein and those made at the hearing, as well as the relevant legal authority, the Court hereby GRANTS IN PART and DENIES IN PART the motion to dismiss, for the following reasons. BACKGROUND1 Lead Plaintiffs Ohio Public Employees Retirement System and PFA Pension Forsikringsaktieselskab’s (collectively, “Plaintiffs”) bring this Private Securities Litigation Reform Act (“PSLRA”) action against Defendant Meta Platforms, Inc. (“Meta”) and Individual Defendants Mark Zuckerberg (Chairman and CEO), David Wehner (Meta’s CFO), Nick Clegg (Meta’s President of Global Affairs), Adam Mosseri (Head of Instagram), Guy Rosen (Meta’s Chief Information Security Officer), Andy Stone (Meta’s Policy Communications Director), Angione Davis (Meta’s Global Head of Security), Karina Newton (Instagram’s Head of Public

1 The Court accepts Plaintiffs’ allegations in the amended complaint as true and construes the Policy), Yann LeCun (Meta’s Chief Artificial Intelligence scientist), Monika Bickert (Meta’s Vice President of Content Policy), and Pavni Diwanji (Meta’s Vice President) (collectively, “Defendants”). In September and October of 2021, the Wall Street Journal published a series of articles entitled the “Facebook Files.” ECF 97 (First Amended Complaint, “FAC”) ¶¶ 1-2. These articles relayed information from whistleblower Frances Haugen, a former Meta data scientist, about Meta’s policies and actions, including how it prioritized revenue at the cost of safety. FAC ¶¶ 1-4. From April 27, 2021, to October 21, 2021, the alleged Class Period, Meta made a series of statements regarding its “cross check” (or “X-check”) or “whitelisting program,” including that it consistently applied its content moderation policies and held everyone to the same standards. FAC ¶¶ 83-86, 90-96. Meta told the public that it has a system known as “cross check” where it gives content from certain newsworthy pages or profiles additional review. FAC ¶ 96. The system exempted millions of influential users from Meta’s Community Standards. FAC ¶¶ 100- 121. In 2018, Meta made changes to its news feed algorithm purportedly to help users have more “meaningful social interactions” or “MSI.” FAC ¶¶ 125-27. Throughout the Class Period, Defendants claimed that Meta’s news feed algorithm and content moderation policies decreased harmful content. FAC ¶¶ 140-159. However, harmful, toxic content and misinformation proliferated. FAC ¶¶ 132-39. During the Class Period, Defendants also reassured the public that Instagram did not cause harm to young users. FAC ¶ 202. They told investors that there was little existing research on the impact of Instagram on children and that any negative effects of the platform were small. FAC ¶ 202. In fact, Meta’s existing internal research showed that Instagram caused or worsened mental health issues for young users. FAC ¶¶ 203, 210-227. Finally, during the Class Period, Meta emphasized its user growth metrics when reporting its daily active users (“DAUs”) and monthly active users (“MAUs”). FAC ¶¶ 258-64. Meta told investors that it regularly evaluated Facebook metrics to estimate the number of “duplicate” or the prevalence of “SUMA” (same user, multiple accounts) accounts. FAC ¶ 270. An internal presentation based on a sample of 5,000 new accounts showed that at least 32% and as many as 56% were opened by existing users. FAC ¶ 270. Plaintiffs bring two causes of action against Defendants for violations of Section 10(b) and Section 20(a) of the Securities and Exchange Act. They allege that Defendants made a series of alleged false or misleading statements regarding four separate aspects of Meta’s business operations: (1) affirming Meta’s commitment to holding all users to the same Community Standards despite Meta’s practice of “cross check” or “X-Check” and “whitelisting” certain individuals, FAC ¶¶ 13-21; (2) misleading investors about Meta’s algorithm and content moderation practices, which encouraged the promotion of divisive content, FAC ¶¶ 22-25; (3) misstatements and omissions about Instagram’s harm to young users, FAC ¶¶ 26-30; and (4) overstating Meta’s growth rate by misleading investors about the number of duplicate accounts among its new users, FAC ¶¶ 31-34. Federal Rule of Civil Procedure 8(a) requires that a complaint contain “a short and plain statement of the claim showing that the pleader is entitled to relief.” A defendant may move to dismiss a complaint for failing to state a claim upon which relief can be granted under Federal Rule of Civil Procedure 12(b)(6). “Dismissal under Rule 12(b)(6) is appropriate only where the complaint lacks a cognizable legal theory or sufficient facts to support a cognizable legal theory.” Mendiondo v. Centinela Hosp. Med. Ctr., 521 F.3d 1097, 1104 (9th Cir. 2008). To survive a Rule 12(b)(6) motion, a plaintiff must plead “enough facts to state a claim to relief that is plausible on its face.” Bell Atl. Corp. v. Twombly, 550 U.S. 544, 570 (2007). A claim is facially plausible when a plaintiff pleads “factual content that allows the court to draw the reasonable inference that the defendant is liable for the misconduct alleged.” Ashcroft v. Iqbal, 556 U.S. 662, 678 (2009). In reviewing the plausibility of a complaint, courts “accept factual allegations in the complaint as true and construe the pleadings in the light most favorable to the nonmoving party.” Manzarek v. St. Paul Fire & Marine Ins. Co., 519 F.3d 1025, 1031 (9th Cir. 2008). Nonetheless, Courts do not unreasonable inferences.” In re Gilead Scis. Secs. Litig., 536 F.3d 1049, 1055 (9th Cir. 2008). Securities fraud cases have heightened pleading requirements as the complaint must satisfy both the pleading requirements of Federal Rule of Civil Procedure 9(b) and the PSLRA. In re VeriFone Holdings, Inc. Sec. Litig., 704 F.3d 694, 701 (9th Cir. 2012). Pursuant to Rule 9(b), claims alleging fraud must “state with particularity the circumstances constituting fraud . . . [.]” Fed. R. Civ. P. 9(b). The PSLRA mandates that “the complaint shall specify each statement alleged to have been misleading, [and] the reason or reasons why the statement is misleading. . . [.]” 15 U.S.C. § 78u–4(b)(1)(B). The PSLRA further requires that the complaint “state with particularity facts giving rise to a strong inference that the defendant acted with the required state of mind.” Tellabs, Inc. v. Makor Issues & Rts., Ltd., 551 U.S. 308, 314 (2007) (quoting 15 U.S.C. § 78u–4(b)(2)(A)). This means a plaintiff must allege that “the defendant[] made false or misleading statements either intentionally or with deliberate recklessness.” In re VeriFone Holdings, 704 F.3d at 701 (quoting Zucco Partners, LLC v. Digimarc Corp., 552 F.3d 981, 991 (9th Cir. 2009)). On January 27, 2023, Defendants filed the instant motion to dismiss. ECF 110. While the scope of review on a motion to dismiss is generally limited to the contents of the complaint, courts may take judicial notice of facts that are “not subject to reasonable dispute.” Fed. R. Evid. 201(b). Courts may consider documents incorporated into the complaint by reference, Tellabs, 551 U.S. at 322, and take judicial notice of documents on which complaints necessarily rely, Lee v. City of Los Angeles, 250 F.3d 668, 688 (9th Cir. 2001), publicly available financial documents such as SEC filings, Metzler Inv. GMBH v. Corinthian Colls., Inc., 540 F.3d 1049, 1064 n.7 (9th Cir. 2008), and publicly available articles or other news releases of which the market was aware, Heliotrope Gen., Inc. v. Ford Motor Co., 189 F.3d 971, 981 n.18 (9th Cir. 1999). Defendants argue that the Court should consider three newspaper articles and posts ECF 110-1. Plaintiffs oppose this request, arguing that Defendants may not use the articles to contradict the complaint. ECF 116. The Court may not assume the truth of an incorporated document “if such assumptions only serve to dispute facts in a well-pleaded complaint.” Khoja v. Orexigen Therapeutics, Inc., 899 F.3d 988, 1003 (9th Cir. 2018). However, the Court may consider the context in which the allegedly false statement was made. See id. at 1002 (holding that incorporation by reference “prevents plaintiffs from selecting only portions of documents that support their claims, while omitting portions of those very documents that weaken – or doom – their claims”); see also In re Intel Corp. Sec. Litig., 2019 WL 1427660, at *6 (N.D. Cal. Mar. 29, 2019) (same). Plaintiffs quote the exhibits extensively in the FAC and allege that each document contained a false or misleading statement. See FAC ¶¶ 90, 419, 475-76, 482-83. As Exhibits 1, 3, and 4 “form the basis of [plaintiffs’] claims,” the Court takes judicial notice of these exhibits. Khoja, 899 F.3d at 1004. Defendants also seek judicial notice of Meta’s Form 10-K for fiscal year 2020. ECF 110-1 at 2. Plaintiffs oppose this request, arguing that the 10-K preceded the Class Period by several months. ECF 116 at 4. However, Plaintiffs cite no authority showing that such timing makes judicial notice improper. Indeed, the Ninth Circuit has repeatedly held that courts may take judicial notice of matters of public record, such as SEC filings. See Metzler, 540 F.3d at 1064 n.7; Dreiling v. Am. Exp. Co., 458 F.3d 942, 946 n.2 (9th Cir. 2006); MGIC Indem. Corp. v. Weisman, 803 F.2d 500, 504 (9th Cir. 1986). The Court accordingly takes judicial notice of the Form 10-K in Exhibit 2. Both parties request that the Court take judicial notice of the Meta Oversight Board’s Policy Advisory Opinion on Meta’s Cross-Check Program for the “limited purpose of establishing that the Oversight Board made the statements contained in the report.” ECF 115 (Ex. A) at 2-3; ECF 121. Courts may take judicial notice of matters of public record and of records and reports of administrative bodies. Khoja, 899 F.3d at 999, 1001. Thus, the Court takes judicial notice of this exhibit. II. SECTION 10(B) CLAIM of Section 10(b) of the Exchange Act and Rule 10b-5. To state a claim under Section 10(b) and Rule 10b-5, a plaintiff must allege “(1) a material misrepresentation or omission by the defendant; (2) scienter; (3) a connection between the misrepresentation or omission and the purchase or sale of a security; (4) reliance upon the misrepresentation or omission; (5) economic loss; and (6) loss causation.” In re Quality Sys., Inc. Sec. Litig., 865 F.3d 1130, 1140 (9th Cir. 2017) (quoting Halliburton Co. v. Erica P. John Fund, Inc., 573 U.S. 258, 267 (2014)); see 17 CFR § 240.10b– 5(b); 15 U.S.C. § 78j(b). “Rule 9(b) applies to all elements of a securities fraud action, including loss causation.” Oregon Pub. Emps. Ret. Fund v. Apollo Grp. Inc., 774 F.3d 598, 605 (9th Cir. 2014). Defendants move to dismiss the Section 10(b) claim for failure to allege (1) a misrepresentation or omission, (2) scienter, and (3) loss causation. The “more exacting pleading requirements” of the PSLRA require that the complaint plead both falsity and scienter with particularity, and Defendants argue that the pleading here falls short of that standard. See Zucco Partners, 552 F.3d at 990; see also 15 U.S.C. § 78u4(b)(1). The Court takes up the three challenged elements in turn. A. Material Misrepresentation or Omission The first issue is whether Plaintiffs sufficiently allege a material misrepresentation or omission. Defendants argue that the challenged statements are not material misrepresentations or omissions because they are (1) opinion statements; (2) nonactionable statements of corporate optimism; and (3) not false or misleading. The Court addresses each argument in turn. 1. Opinion Statements Meta argues that statements of opinion about its content moderation efforts are not actionable because Plaintiffs have not alleged that they are both subjectively and objectively false. Mot. at 30 (citing FAC ¶¶ 426, 424, 427, 437, 439, 442). Expressions of opinions – as opposed to statements of fact – are only actionable if they are both subjectively and objectively false or misleading. Rubke v. Capitol Bancorp Ltd, 551 F.3d 1156, 1162 (9th Cir. 2009) (citing Va. Bankshares, Inc. v. Sandberg, 501 U.S. 1083, 1094-96 (1991)). “To be misleading, a statement Fund v. Hewlett-Packard Co., 845 F.3d 1268, 1275 (9th Cir. 2017) (quoting Oregon Pub., 774 F.3d at 606). In several opinion statements (Statements 19, 20, 25, 27)2 that Plaintiffs offer in support of their Section 10(b) claim, they fail to allege that the statements were objectively and subjectively false or misleading. For example, in Statement 19, Defendant Wehner states “I think more than anyone else in the industry, we invest on the safety and security side to sort of keep bad content off the site . . . [.]” FAC ¶ 426. Plaintiffs fail to allege any facts showing that this statement is objectively false – i.e., that Meta does not do more to prevent misinformation than others in the industry. Plaintiffs also fail to allege facts showing that Wehner did not believe this statement to be true. Accordingly, Statement 19 is not actionable. Similarly, Plaintiffs fail to plead that Statement 20 was objectively or subjectively false when made. In Statement 20, Zuckerberg stated that Meta announced a “big shift” in 2018 “knowing that people would spend less time on Facebook” to help that have more “meaningful social interactions,” and that Meta viewed the shift “success because it improves the experience of our users.” Statement 20 (FAC ¶ 427). Plaintiffs fail to allege that this statement was false or that Zuckerberg did not believe it to be true. The Court therefore dismisses the claims based on Statements 20 on this basis as well. However, the Court cannot agree that Statements 25 or 27 are opinion statements because Defendants make factual assertions in these statements: • “You may claim [Facebook] amplifies ideas you disagree with, but it’s simply false.” Statement 25 (FAC ¶ 437). • “The idea that there is always a causal link between social media and polarization is not true . . . the research [shows that] because most people. . . tend to have a more heterogenous and mixed ideological composition of [] friends on Facebook than you do if you read the same partisan newspaper or watch the same cable TV or news outlet,” social media is “not as narrow in terms of its ideological diet.” Statement 27 (FAC ¶ 442).

2 The Court refers to the statements as they are numbered (1-60) in the Appendix submitted by the These statements assert that Facebook does not amplify ideas that users disagree with and that there is no causal link between social media and polarization. As these statements are “capable of objective verification,” Retail Wholesale, 845 F.3d at 1275 (citation omitted), the Court cannot dismiss them as nonactionable opinion statements.3 2. Corporate Puffery Defendants also argue that several statements (Statements 2, 4, 7, and 16, 17, 19, 21-24, 26, 27, and 33) regarding Meta’s enforcement of community standards and its algorithm and content moderation efforts are nonactionable statements of corporate optimism. “[V]ague, generalized assertions of corporate optimism or statements of ‘mere puffing’ are not actionable material misrepresentations under federal securities laws” because no reasonable investor would rely on such statements. City of Royal Oak Ret. Sys. v. Juniper Networks, Inc., 880 F. Supp. 2d 1045, 1063 (N.D. Cal. 2012) (citations omitted). This is because “[w]hen valuing corporations, . . . investors do not rely on vague statements of optimism like ‘good,’ ‘well-regarded,’ or other feel good monikers.” In re Cutera Sec. Litig., 610 F.3d 1103, 1111 (9th Cir. 2010). However, “even ‘general statements of optimism, when taken in context, may form the basis for a securities fraud claim’ when those statements address specific aspects of a company’s operation that the speaker knows to be performing poorly.” Quality Sys., 865 F.3d at 1143 (citation omitted). “Statements by a company that are capable of objective verification are not “puffery” and can constitute material misrepresentations.” Oregon Pub., 774 F.3d at 606. Defendants argue that Statements 19, 22, and 33 about Meta’s algorithm and content moderation efforts are nonactionable statements of corporate optimism: • “I think more than anyone else in the industry, we invest on the safety and security side to sort of keep bad content off the site before it gets ranked and put into what people see. . . . I think we really do more than anyone else in the industry on the safety and security front to prevent things like misinformation and bad content going into the

3 Defendants also argue that Statements 18 (FAC ¶ 424) and 26 (FAC ¶ 439) are non-actionable opinion statements. As Defendants analyze whether the statements are false or misleading along system in the first place.” Statement 19 (FAC ¶ 426).4 • “We’re taking significant steps to fight the spread of misinformation . . . [.]” Statement 22 (FAC ¶ 429).5 • “[F]or the last several years, our team that has been focused on kind of trust and safety overall, has been more focused on content moderation, so making sure that we can identify harmful content and take it down and also working on making sure that we continue improving privacy and supporting and providing world-class controls there.” Statement 33 (FAC ¶ 457). Statements that Meta takes “significant steps” to fight disinformation, that it believes it does more than others in the industry on safety and security, and is focused on “trust and safety” and improving privacy are too generalized and do not describe a “specific and testable characteristic” of Meta’s platform. See Andersen v. Griswold Int’l, LLC, 2014 WL 12694138, at *5 (N.D. Cal. Dec. 16, 2014); see e.g., Lloyd v. CVB Fin. Corp., 811 F.3d 1200, 1206-07 (9th Cir. 2016) (concluding that “strong credit culture and underwriting integrity remain paramount at CVB” was non-actionable puffery); In re Yahoo! Inc. Customer Data Sec. Breach Litig., 2017 WL 3727318, at *26 (N.D. Cal. Aug. 30, 2017) (concluding that “protecting our system and our users’ information is paramount to ensure Yahoo users enjoy a secure user experience and maintaining our users’ trust” was non-actionable puffery as it did not state any specific characteristics of Defendants’ products or services). As Statements 19, 22, and 33 are too vague and generalized to be actionable, the Court dismisses the claims based on those statements. See Baltazar v. Apple Inc., 2011 WL 6747884, at *4 (N.D. Cal. Dec. 22, 2014) (holding that “[i]f an alleged misrepresentation would not deceive a reasonable consumer or amounts to mere puffery, then the claim may be dismissed as a matter of law.”).

4 Although the Court already found that Statement 19 is an opinion statement that Plaintiff failed to adequately plead, it also finds that this statement is corporate puffery.

5 This statement also proclaims that “[o]ur machine learning models to find potentially violating COVID-19 and vaccine content are trained to surface content in 19 languages.” However, Plaintiffs have not offered any allegations showing that this portion of the statement is false or However, the Court cannot agree that the additional statements classified by Defendants as puffery (Statements 2, 4, 7, and 16, 17, 21, 23, 24, 26, and 27) are too vague or generalized to be actionable. “[T]here is a difference between enthusiastic statements amounting to general puffery and opinion-based statements that are anchored in misrepresentations of existing facts.” City of Sunrise Firefighters’ Pension Fund v. Oracle Corp., 527 F. Supp. 3d 1151, 1172 (N.D. Cal. 2021) (internal quotation marks and citations omitted). For example, in Statement 2, Meta told the Board that “it applies a ‘cross check’ system to some ‘high profile’ accounts to ‘minimize the risk of errors in enforcement’ . . . [but] ‘it has never had a general rule that is more permissive for content posted by political leaders.’ ” Statement 2 (FAC ¶ 389). Meta also stated that it removes content that violates its standards and does not have special protections for any groups. Statement 4 (FAC ¶ 393); Statement 16 (FAC ¶ 419). These statements suggest specific actions that Meta took and rules that it implemented or enforced. Thus, these statements can be verified and are not corporate puffery. See In re Cutera, 610 F.3d at 1111. While portions of Statements 4 and 7 relate to Meta’s vague goals and aspirations,6 Defendants fail to address the remaining portions of the statements, which state more concrete facts. In Statements 4 and 7, Defendants claim that they remove all content that violates Meta’s Community Standards, and that Meta assesses everyone under the same Community Standards. Statement 4 (FAC ¶ 393); Statement 7 (FAC ¶ 399). Defendants do not explain why these statements are not capable of objective verification and the Court cannot find that the statements are corporate puffery. Similarly, the Court cannot agree that Statements 17, 18, 21, 23, 24, 26, and 27 fail as too vague or general. Meta argues only that these statements are “subjective assessments” of the content Meta wants on its apps or what is in its financial or reputational interest. Mot. at 29. However, the statements assert that Meta removes certain content from its platform, has reduced the prevalence of disinformation, and that it is not in Facebook’s interest to push users towards

6 See FAC ¶¶ 393, 399 (Meta “strive[s] to enforce our policies evenly”); Klein v. Facebook, Inc., 580 F. Supp. 3d 743, 795 (N.D. Cal. 2022) (finding the statement “[k]eeping the global extreme content. Such assertions are objectively verifiable. See In re Yahoo!, 2017 WL 3727318, at *26 (finding that “physical, electronic, and procedural safeguards that comply with federal regulations to protect personal information about you” is not puffery as it made a specific guarantee that Defendants “use safeguards that complied with federal regulations to protect users’ information”). Therefore, the Court cannot dismiss the claims based on Statements 2, 4, 7, 16-18, 21, 23, 24, 26, and 27 as on this basis. To recap, the Court finds that Statements 19, 22, and 33 are not actionable statements of corporate puffery and Plaintiffs failed to allege that Statement 22 is objectively and subjectively false or misleading. Statements 1-18, 21, 23-32, and 34-60 remain live. 3. False or Misleading The Court next addresses whether Plaintiffs have sufficiently alleged that the remaining 56 statements (Statements 1-18, 21, 23-32, 34-60) are false or misleading. “For a statement to be false or misleading, it must ‘directly contradict what the defendant knew at that time’ or ‘omit[] material information.’ ” Weston Fam. P’ship LLLP v. Twitter, Inc., 29 F.4th 611, 619 (9th Cir. 2022) (quoting Khoja, 899 F.3d at 1008-09). Statements that are “literally true” may still be misleading due to “their context and manner of presentation.” Miller v. Thane Int’l, Inc., 519 F.3d 879, 886 (9th Cir. 2008). “The most direct way to show both that a statement was false when made and that the party making the statement knew that it was false is via contemporaneous reports or data, available to the party, which contradict the statement.” Nursing Home Pension Fund Local 144 v. Oracle Corp., 380 F.3d 1226, 1230 (9th Cir. 2004). The parties categorize the challenged statements in four categories: (a) “Cross Check” or enforcement of community standards; (b) algorithm and content moderation efforts; (c) Instagram’s impact on young users; and (d) the prevalence of duplicate accounts. The Court analyzes whether the challenged statements are false or misleading in each of the four categories. a. Cross Check: Enforcement of Community Standards Plaintiffs allege 16 statements (Statements 1-16) that Defendants made about cross-check throughout the Class Period. Meta’s “cross-check” or “X-check” program exempts influential content posted by VIPs rather than allowing “newsworthy” content to remain online. FAC ¶¶ 100- 115, 403; see FAC ¶ 102 (cross check “shield[ed] millions of VIP users from the company’s normal enforcement process”). They allege that the following statements are false or misleading: Meta applies its community standards and policies to everyone and all types of content (Statements 1, 3-5, 7, 12, and 13); Meta removes harmful content no matter who posts it (Statements 6, 10, 13, and 14); cross-check is simply a system of additional review for certain profiles or pages (Statements 9, 15, and 16); and cross-check applies in limited newsworthy circumstances (Statements 2, 8, and 11). Defendants move to dismiss claims based on these statements, arguing that Plaintiffs do not provide particularized allegations contradicting the alleged misstatements. Mot. at 23-24. They argue that Plaintiffs cite “isolated sentences, devoid of context” from unsourced internal Meta documents from 2019 and 2020, a year to two years before the class period. Id. To show that a statement was false when made, a plaintiff may point to “contemporaneous reports or data.” See Nursing Home, 380 F.3d at 1230. Where a complaint relies on internal reports, it must contain “at least some specifics from those reports as well as such facts as may indicate their reliability.” Lipton v. Pathogenesis Corp., 284 F.3d 1027, 1036 (9th Cir. 2002) (citation omitted). In the FAC, Plaintiffs quote Wall Street Journal articles which reported that internal Meta documents showed that “most of the content flagged by the cross-check system did not face subsequent reviews,” and by 2020, at least 5.8 million users were on the cross-check list. FAC ¶ 103. According to a 2019 internal audit, at least 45 teams at Meta were involved in whitelisting, many of which “chose not to enforce the rules with high profile accounts at all.” FAC ¶ 104. A 2019 audit found that Meta “currently review[s] less than 10% of XChecked content.” FAC ¶ 106. Plaintiffs fail to allege specific facts to overcome the heightened pleadings requirements of the PSLRA and Rule 9(b). In re Verifone Holdings, 704 F.3d at 701. Plaintiffs argue that the quotations from internal Meta documents and whistleblower testimony “surpass[] the level of detail required by the PSLRA.” ECF 113 (“Response”) at 21. However, Plaintiffs plead selective quotes from unidentified internal Meta documents that the Wall Street Journal reported. See reviewed them, or what the reports entailed. Essentially, Plaintiffs fail to state with particularity the needed “who, what, when, where, and how of the misconduct charged . . . [.]” Davidson v. Kimberly-Clark Corp., 889 F.3d 956, 964 (9th Cir. 2018); see In re Silicon Graphics Inc. Sec. Litig., 183 F.3d 970, 985 (9th Cir. 1999) (finding that the complaint did not contain “adequate corroborating details” where plaintiff did not provide an “adequate description of [the] contents” of reports or “facts as may indicate their reliability”).7 Accordingly, the Court dismisses the claim based on Statements 1, 2, 3-7, 9, 12-14, 15, and 16. With respect to the statements that cross-check only applied to a limited number of posts, Plaintiffs also point to an October 2021 transparency report published by Meta’s Oversight Board. That report acknowledged that it was misleading to state that cross-check applied to only a “small number of decisions.” FAC ¶ 119. The report indicates that Meta had not been “fully forthcoming in its responses on cross-check.” FAC ¶¶ 119-20. Considering the 2019 and 2020 reports in tandem with the 2021 transparency report, Plaintiffs have sufficiently alleged that statements that cross-check only applied in limited newsworthy situations (Statements 8 and 11) were misleading. See Nursing Home, 380 F.3d at 1230 (holding that the “most direct way to show both that a statement was false when made and that the party making the statement knew that it was false is via contemporaneous reports or data, available to the party, which contradict the statement”). Therefore, the claim based on Statements 8 and 11 stand. b. Algorithm and Content Moderation Efforts Plaintiffs next challenge 23 statements (Statements 17-39) that Defendants made about Meta’s algorithm and content moderation. Throughout the Class Period, Meta stated that it proactively removed hate speech and misinformation from its platforms, did not profit from polarization or prioritize profit over safety, and that hate speech prevalence on Facebook

7 Statement 2 was made to the Meta Oversight Board in response to questions it asked Meta. FAC ¶ 389. In order to be actionable, a statement must be made “in a manner reasonably calculated to influence the investing public.” McGann v. Ernst & Young, 102 F.3d 390, 393 (9th Cir. 1996) (citations omitted). Plaintiffs have not alleged that Meta’s statement to its Board was made in such a manner. Moreover, even if the statement were otherwise actionable, Plaintiffs have failed to allege who made these statements and when. See Davidson, 889 F.3d at 964. decreased due to changes it made to the platform. FAC ¶¶ 422, 424, 426-430, 435, 437, 439, 442, 445, 447, 450, 451, 454, 457, 459, 461, 463, 465-467. Plaintiffs allege that these statements are materially false or misleading because Meta’s algorithm amplified harmful content “by design,” Meta’s content-moderation tools did not sufficiently address harm on the platform, most hate speech removed was reported by the community, not removed proactively, and Meta profited from harmful content. Response at 26-27. Meta argues that none of the statements were false or misleading, addressing algorithm and content moderation separately. Following the motion’s convention, the Court first addresses whether Plaintiffs have adequately alleged that the algorithm statements are false or misleading and then takes up the content moderation statements. i. Algorithm Statements The challenged algorithm statements (Statements 17-18, 21, 23-25) are: • “There are certain types of content we simply don’t allow on our services.” Statement 17 (FAC ¶ 422). • “[W]e don’t want [] extremist content or any of that stuff on our services . . . we go out of our way to try to reduce it.” Statement 18 (FAC ¶ 424). • The algorithm “reduce[d] the distribution of many types of content.” Statement 21 (FAC ¶ 428). • Meta “remove[s] information that contributes to the risk of imminent violence or physical harm” and “introduced tools that allow us to proactively detect and remove violating content.” Statement 23 (FAC ¶ 430) • Meta’s investment in AI has made its “spam detection and [] integrity systems more effective at identifying stuff.” Statement 24 (FAC ¶ 435) • It is “simply false” that Facebook “amplifies ideas you disagree with.” Statement 25 (FAC ¶ 437). Plaintiffs argue that these statements are false and misleading because the algorithm amplified toxic content, Meta profited from polarization, its systems did not sufficiently address harm, and Meta took action on very little hate content. Response at 29. Plaintiffs point to the Wall Street expressing “anger and outrage was far more likely to score higher and was weighted more heavily by the algorithm.” FAC ¶ 135; see FAC ¶ 136. Plaintiffs also cite an internal Meta research report from the Journal article titled “What is Collateral damage?” that concluded based on “compelling evidence” that the algorithm itself led to the proliferation of harmful content. FAC ¶ 167; see also FAC ¶¶ 166, 188, 568 (internal Meta research and reports from 2018 and 2019 concluding that Meta’s approach has “unhealthy side effects on important slices of public content” and that content moderation was not sufficient to address harm on the platform). In sum, Plaintiffs allege that the Journal quoted internal Meta reports or memos from 2017-2019 that concluded the Facebook algorithm had a side effect of causing harmful content to proliferate. However, as with the cross-check statements, Plaintiffs selectively quote from Meta reports referenced in the Journal article without “adequate corroborating details” of the contents of the reports that would indicate their reliability, such as who wrote the reports, or what research or content they contained. See In re Silicon Graphics, 183 F.3d at 985 (finding that complaint failed to provide “specifics” from internal reports such as the sources of information, who drafted the reports, or which officers received them). Further, in relying on reports from 2017, 2018, and 2019, Plaintiffs do not show that the challenged statements made in 2021 were false when made. Nursing Home, 380 F.3d at 1230 (requiring that a statement was false “when made” based on “contemporaneous” reports or data); Crews v. Rivian Auto., Inc., 2023 WL 4361098, at *8 (C.D. Cal. July 3, 2023) (same). Plaintiffs’ allegations are thus insufficient to show that the statements about the algorithm were false when made. See In re Syntex Corp. Sec. Litig., 95 F.3d 922, 930 (9th Cir. 1996) (finding that internal memorandum written months after statement did not make memorandum false). Even if the Court were to consider the reports, Plaintiffs offer no factual allegations that contradict Meta’s statements that it was taking significant steps to fight misinformation (Statement 22, FAC ¶ 429), that Meta removes misinformation from its platforms (Statement 23, FAC ¶ 430), or that Meta’s investment in artificial intelligence makes its spam detection more effective (Statement 24, FAC ¶ 435). Additionally, the conclusion in the reports that the algorithm had negative side effects does not contradict Defendants’ assertion that they that it is not in their interest to push users toward such content (Statements 17, 18, 21, 22, 23). The Court therefore dismisses the claims based on Statements 17-18 and 21-25. ii. Content Moderation Statements Plaintiffs also allege that Defendants made materially false and misleading statements about its content moderation efforts (Statements 26-39), including: • “[W]e catch the vast majority of [hate speech] before anyone reports it to us . . . the idea that we have an incentive to prioritize this, is I think one of the most misleading allegations . . . [.]” Statement 26 (FAC ¶ 439); • “The idea that there is always a causal link between social media and polarization is not true . . . the research [shows that] because most people. . . tend to have a more heterogenous and mixed ideological composition of [] friends on Facebook than you do if you read the same partisan newspaper or watch the same cable TV or news outlet,” social media is “not as narrow in terms of its ideological diet.” Statement 27 (FAC ¶ 442); • Meta banned over 250 white supremacist groups from its platform and continued to enforce its ban on hate groups. Statement 28 (FAC ¶ 445); • Meta has “no incentive [] to do anything but remove . . . language that incites or facilitates violence” and bans hate groups. Statement 29 (FAC ¶ 447); • Meta tries to “keep the prevalence of hate speech on our platform as low as possible” and the “improvement in prevalence on Facebook is due to changes we made to reduce problematic content in News Feed. . . . Today we proactively detect about 97% of hate speech content we remove.” Statement 30 (FAC ¶ 450); • “[H]ate speech prevalence on Facebook continues to decrease” due to changes Meta “continue[s] to make to reduce problematic content in the newsfeed.” Statement 31 (FAC ¶ 451); • A proposal requiring Meta to take additional steps to address hate speech and violent content on its platform was “unnecessary and not beneficial” to shareholders as Meta already removes information that violates its Community Standards on violence, bullying, • Meta is “making sure fewer people see misinformation on our apps” and has been focused on “reducing viral misinformation.” Statement 34 (FAC ¶ 459); see Statement 33 (FAC ¶ 457) (Meta has focused on identifying harmful content to be able to take it down). • Meta removed “31.5 million pieces of hate speech content from Facebook, compared to 25.2 million in Q1, and 9.8 million from Instagram, up from 6.3 million in Q1.” Statement 35 (FAC ¶ 461); see Statement 36 (FAC ¶ 463); Statement 39 (FAC ¶ 467). • “At the heart of [the Wall Street Journal] series is an allegation that is just plain false: that Facebook conducts research and then systematically and willfully ignores it if the findings are inconvenient for the company.” Statement 37 (FAC ¶ 465). • “We do not profit from polarization, in fact, just the opposite.” Statement 38 (FAC ¶ 466). Plaintiffs argue that these statements touting Meta’s content moderation practices are false or misleading because Meta “estimate[s] that [it] may take action on as little as 3-5% of hate and ~0.6% of V&I [violence and incitement] on Facebook despite being the best in the world at it,” hate speech was “inordinately prevalent,” content moderations practices were “not remotely sufficient,” and “98 per cent of the Hate Speech contents are reported reactively.” Response at 29- 30 (citing FAC ¶¶ 124, 79-88, 192-93). Defendants argue that Plaintiffs fail to allege that any of the content moderation statements were false or misleading. Mot. at 32-33. The Court agrees. First, Plaintiffs have failed to point to factual allegations in their complaint that it was false or misleading for Meta to state that it banned hate groups (Statement 28), profited from – or had an incentive to promote – polarization (Statements 29, 38), or that friends on Facebook are less heterogeneous than a partisan newspaper or TV show (Statement 27). Accordingly, the Court dismisses the claims based on Statements 27, 28, 29, and 38. Second, Plaintiffs rely on an undated internal analysis of “Hate Speech break-down by detection source,” which found that 98% of hate speech is reported reactively by the community and only two percent is caught proactively. FAC ¶ 193. However, Plaintiffs fail to allege when this report was made, by whom, or what data it analyzed. See Lipton, 284 F.3d at 1036 (holding that “negative characterizations of reports relied on by insiders, without specific reference to the PSLRA”). Thus, Plaintiffs fail to allege that Statements 26 and 30, which discuss Meta proactively detecting hate speech, are false or misleading. Third, Plaintiffs allege that a March 2021 report indicates that Meta may take action on “as little as 3-5% of hate and ~0.6% of [violence and inciting content] on Facebook despite being the best in the world at it.” FAC ¶¶ 193, 343. Defendants argue that this quote from an unsourced document fails to contradict Meta’s statements about its content moderation efforts (and praises Meta as “the best in the world” at removing content). Mot. at 33. As above, the selective quotation fails to provide any specific factual allegations about the report, such as who drafted it and any specifics from the report (e.g., whether it analyzed all violent and inciting content on Facebook or a specific region). See Lipton, 284 F.3d at 1036 (requiring a complaint relying on internal reports to contain “at least some specifics” from the reports to indicate their reliability). Nevertheless, even if the Court were to consider these allegations, they fail to contradict Meta’s statements that it focuses on reducing and removing misinformation and hate content or that the prevalence of such speech has declined. Accordingly, the Court dismisses the claims based on Statements 31-39. Plaintiffs have failed to allege that any of the algorithm or content moderation statements are false or misleading. c. Instagram’s Impact on Young Users Plaintiffs next challenge 17 statements that Defendants made about Instagram’s impact on young users (Statements 40-56). In sum, Meta made statements that 1) there is little research on Instagram’s impact on children; 2) Instagram’s impact on the mental health of teens is bi- directional; 3) social media is not inherently bad for teens and children; and 4) Meta is committed to protecting child safety. Plaintiffs allege that Defendants’ statements were false because Meta conducted internal research over several years that Instagram “makes teens feel very bad,” negative effects of Instagram on teens were “common” and “severe,” made “body images worse for 1 in 3 teenage girls,” and 14% of boys in the U.S. reported that Instagram made them feel worse about themselves. FAC ¶¶ 203, 210-227. (Statements 40, 42, and 50) are not actionable as they were not made “in connection with the purchase or sale of any security.” Mot. at 37 (citing 15 U.S.C. § 78j(b)). The Court declines to dismiss these statements on this basis. The cases cited by Defendants “do not indicate that only market-related documents, such as regulatory filings, public presentations, or press releases, can contain actionable misstatements under Section 10(b).” In re Volkswagen “Clean Diesel” Mktg., Sales Pracs., & Prod. Liab. Litig., 2017 WL 66281, at *18 (N.D. Cal. Jan. 4, 2017) (emphasis in original); In re Tesla, Inc. Sec. Litig., 477 F. Supp. 3d 903, 922-27 (N.D. Cal. 2020) (finding that complaint alleged that Musk’s Tweets and blog posts were material misrepresentations). At this stage, the Court cannot declare as a matter of a law that a reasonable investor would not have considered statements made by the head of Instagram on his personal Instagram account or an interview, or statements on the official blog on Meta’s website, to be material investing information. See, e.g., TSC Indus., Inc. v. Northway, Inc., 426 U.S. 438, 450 (1976) (assessing materiality is “peculiarly one[] for the trier of fact”). Accordingly, the Court turns to Defendants’ arguments that Plaintiffs failed to allege that the statements about Instagram were false or misleading. Plaintiffs challenge several statements in the FAC that they do not allege were false or misleading. For example, in response to a question about whether Defendant Mosseri (Head of Instagram) believes social media is good for children under 13, he responded that there is “little existing research.” Statement 40 (FAC ¶ 471). In the FAC, Plaintiffs allege that Meta conducted research on the impact of social media on teenagers, see FAC ¶¶ 210-222, but they do not allege that Meta researched its effects on children under 13 years old. Plaintiffs argue that Defendants use “teens,” “tweens,” “children,” and “YA” interchangeably. Response at 39 (citing FAC ¶¶ 234-39). The cited paragraphs in the FAC do not support that assertion because nothing in the FAC has alleged that Meta conducted or had access to research on children under 13. Therefore, Plaintiffs have not alleged that Statement 40 is false or misleading. Similarly, Plaintiffs have not alleged that Statement 42 is false or misleading. In Statement 42, Mosseri stated that Meta’s research on wellbeing was inconclusive because “[w]ellbeing is (FAC ¶ 475). As this statement is an opinion, it must be both objectively and subjectively false. Rubke, 551 F.3d at 1162. Plaintiffs have not alleged that this statement is objectively false or that Mosseri did not believe that wellbeing is “hard to measure” or “a bit subjective.” Accordingly, the Court dismisses the claim based on Statement 42. Similarly, in Statement 50, Meta made a blog post – after the corrective disclosures in the Journal article – stating generally that “social media isn’t inherently good or bad for people.” Statement 50 (FAC ¶ 492). Plaintiffs have not alleged that this statement is objectively verifiable.8 This statement is not actionable. The Court likewise finds that Plaintiffs have not alleged that Statement 44 is false or misleading. There, Mosseri stated that “concerns about Facebook’s overall impact on its users’ well-being are likely overblown.” FAC ¶ 479. Although there are extensive allegations about Meta’s internal research on the harm caused by Instagram, Plaintiffs have not pointed to any allegations about the effect of Facebook on well-being. See FAC ¶ 480. Accordingly, the Court dismisses Statement 44. However, Plaintiffs have sufficiently alleged falsity as to Statements 41, 43, 45-48, and 51- 56. In response to questions about the impact of Instagram on the mental health of teenage girls, Mosseri stated that that research findings are “bi-directional, so small effects positive and small effects negative but it’s quite small.” Statement 45 (FAC ¶ 482). Defendants argue that this is a statement of opinion. However, statements about research findings that do not use “opinion- qualifying language” such as “I think” or “I believe” are statements of fact. In re QuantumScape Sec. Class Action Litig., 580 F. Supp. 3d 714, 739 (N.D. Cal. 2022) (finding that statements about what “data demonstrat[es]” and “testing showed” were not opinions). As a statement of fact, the Court must determine whether Plaintiffs sufficiently allege that Statement 45 is false or misleading. Plaintiffs allege that Meta’s internal research demonstrates that Instagram harms

8 Moreover, Meta made this statement after the corrective disclosure about Instagram’s impact on teens had already been made public. Thus, Plaintiffs cannot allege a “causal connection between the [Defendant’s alleged] material misrepresentation and the [Plaintiffs’ alleged economic] loss.” young users. Response at 40 (citing FAC ¶¶ 203, 210-27). Defendants argue that Plaintiffs cherry-pick the documents they rely on and that the FAC also cites to positive external research showing the benefits of social media. Mot. at 39; see FAC ¶ 300. However, the external research that the FAC cites refers broadly to social media and does not specifically reference Instagram or Meta’s platforms. FAC ¶ 300. Given the extensive internal research that Plaintiffs cite showing the harm that Instagram causes to teenagers, they have sufficiently alleged that stating that research findings are “bi-directional” was false or misleading. Similarly, stating that the research showed that teenage girls said Instagram was helping their mental health, Statement 52 (FAC ¶ 495), was misleading given Meta’s extensive internal research contradicting this assertion. Defendants also contend that Plaintiffs have not alleged that Meta’s various statements about its commitment to keeping children safe are false or misleading.9 Mot. at 40; see, e.g., Statement 46 (Meta has “robust policies to help protect against child exploitation and content or behavior on our platform that puts the safety of children at risk”); Statement 41 (“We remove content that encourages suicide or self-injury on Facebook and Instagram”); Statement 47 (“We want young people to enjoy using Instagram while making sure we never compromise on their privacy and safety”). Plaintiffs allege that these statements misleadingly omitted material facts that Meta documented severe harm to teens on Instagram and Defendants profited from Instagram’s harm to children. Response at 37, 40; see, e.g., FAC ¶ 219 (internal 2019 study showing that teen girls said that Instagram made thoughts of suicide and self-injury worse); FAC ¶ 414 (internal memo showing that Meta was cautious about statements against human trafficking to avoid “alien[ating] buyers” and was “more often concerned with retaining users . . . than it was with preventing human trafficking on its platforms”). Defendants point to no caselaw or authority showing why these allegations are insufficient to allege that the statements are false or misleading. Defendants have not carried their burden, and the Court accordingly declines to dismiss the claims based on Statements 41, 46, and 47. Defendants also challenge claims based on Statement 43 (Meta “has been working closely

9 Defendants argue that these statements are too general to be a ground for securities fraud. As with third-party experts to better understand how to empower people, build self-awareness and shape a more positive experience on Instagram”) and Statement 48 (Meta “convened a group of experts in the fields of online safety, child development and children’s media to share their expertise, research and guidance”). They argue that internal research about Instagram does not evidence that these statements were false. Mot. at 39-40. However, Plaintiffs allege that at the time these statements were made, Meta had already concluded that its proposed measures did not improve well-being but that touting them “could make them look good.” FAC ¶ 478. Accordingly, Plaintiffs sufficiently alleged that these statements are misleading, and Defendants have not met their burden of showing that claims based on these statements should be dismissed. Finally, Defendants challenge the remaining statements (Statements 51 and 53-56) in argument in a footnote. Mot. at 40 n.6. The Court does not address arguments relegated to footnotes. See Est. of Saunders v. Comm’r, 745 F.3d 953, 962 n.8 (9th Cir. 2014) (holding that “[a]rguments raised only in footnotes, or only on reply, are generally deemed waived.”). Accordingly, Plaintiffs have sufficiently alleged that Statements 41, 43, 45-49, 51-56 are false or misleading.10 d. Prevalence of Duplicate Facebook Accounts Finally, Plaintiffs challenge four statements about Facebook’s growth metrics and the prevalence of the same users with multiple accounts (“SUMA”): • Meta reported favorable user growth rates for the first two quarters of 2021: “Facebook’s daily active users (DAUs) were 1.88 billion on average for March 2021, an increase of 8% year-over-year,” “monthly active users (MAUs) were 2.85 billion as of March 31, 2021, an increase of 10% year-over-year,” daily active people (DAP) increased 15%, family monthly active people (MAP) increased 15%, and worldwide DAUs increased 8% from March 2020 to March 2021. Statement 57 (FAC ¶ 504); see also Statement 59 (FAC ¶ 508) (reporting numbers and percentage change for June 2021). • “Duplicate and false accounts are very difficult to measure at our scale, and it is possible that the actual number of duplicate and false accounts may vary significantly from our estimates. In the fourth quarter of 2020, we estimated that duplicate accounts may have represented approximately 11% of our worldwide MAUs. We believe the percentage of duplicate accounts is meaningfully higher in developing markets such as the Philippines and Vietnam, as compared to more developed markets.” Statement 58 (FAC ¶ 506); see Statement 60 (FAC ¶ 510) (same). Plaintiffs allege that these statements were false or misleading because of the prevalence of duplicate accounts, which artificially inflated growth rates. The basis for this assertion is an internal Meta presentation from Spring 2021 that found that “the phenomenon of single users with multiple accounts” was “‘very prevalent’ among new accounts,” and that “of roughly 5,000 recent sign-ups on the service [] at least 32% and as many as 56% were opened by existing users.” FAC ¶ 370. An internal memo from May 2021 stated that “the number of U.S. Facebook users who are in their 20s and active at least once a month often exceeds the total population of Americans their age,” and that the active user numbers were “less trustable.” FAC ¶ 371. Defendants argue that these statements are not false or misleading because Meta did not have a duty to disclose an estimate of duplicate accounts among new accounts. “Rule 10b–5 prohibits ‘only misleading and untrue statements, not statements that are incomplete.’ ” Police Ret. Sys. of St. Louis v. Intuitive Surgical, Inc., 759 F.3d 1051, 1061 (9th Cir. 2014) (quoting Brody v. Transitional Hosps. Corp., 280 F.3d 997, 1006 (9th Cir. 2002)). That is, the Ninth Circuit has “expressly declined to require a rule of completeness for securities disclosures because ‘[n]o matter how detailed and accurate disclosure statements are, there are likely to be additional details that could have been disclosed but were not.’ ” Id. (quoting Brody, 280 F.3d at 1006). However, a plaintiff can still state a 10(b) claim “based on a failure to provide ‘context’ where that failure “affirmatively create[s] an impression of a state of affairs that differs in a material way from the one that actually exists.” Shenwick v. Twitter, Inc., 282 F. Supp. 3d 1115, 1136 (N.D. Cal. 2017) (quoting Brody, 280 F.3d at 1006). Meta’s internal research. First, the presentation on which Plaintiffs rely to show that the growth rates were misleading show only that “of roughly 5,000 recent sign-ups . . . at least 32% and as many as 56% were opened by existing users.” FAC ¶ 370. Plaintiffs have not alleged that Meta extrapolated this small sample to all new accounts or that it can be so generalized. Mot. at 43-44. Plaintiffs rely heavily on Shenwick, 282 F. Supp. 3d at 1137. There, Twitter reported positive MAU growth but was simultaneously experiencing flat or declining DAU trends and other problems with user engagement that made their MAU numbers implausible. Id. Twitter challenged plaintiffs’ metrics and comparison of the ratios. Id. However, Twitter admitted that the DAU to MAU ratio had fallen, and plaintiffs alleged a decline in overall DAU. Id. at 1137-38. In addition, plaintiffs’ allegations were consistent with confidential witness testimony, and Twitter compared DAU to MAU ratios for its top markets to extrapolate DAU growth and criticized plaintiffs for the same type of comparison. Id. Plaintiffs here have not offered anything comparable. Indeed, the allegations here only indicate that a small sample size showed a large percentage of duplicate sign-ups. Plaintiffs have thus failed to show that this presentation made Meta’s statements about DAUs and MAUs (Statements 57 and 59) false or misleading because Plaintiffs have not alleged who created this presentation, or that it applied broadly to DAUs and MAUs. Nor do they allege that it was false or misleading for Meta to state that duplicate accounts are “very difficult to measure at our scale” and that the actual numbers “may vary significantly from our estimates.” Statement 58 (FAC ¶ 506); see Statement 60 (FAC ¶ 510) (same). Moreover, Plaintiffs do not explain why Meta’s disclosure that it believes that the percentage of duplicate accounts is meaningfully higher in developing markets is false or misleading. In Statements 58 and 60, Plaintiffs allege that Defendants’ estimate that duplicate accounts represented approximately 11% of worldwide MAUs in the fourth quarter of 2020 was misleading. FAC ¶¶ 506, 510. However, Plaintiffs point to no allegations explaining why this was misleading or that Defendants knew that the percentage was different. Accordingly, the Court finds that Plaintiffs have not alleged that any of the statements regarding growth rates (Statements 57-60) were false or misleading, and dismisses the claims based on these statements. B. Scienter As detailed above, Plaintiffs have sufficiently alleged that Statements 8 and 11 (cross- check) and Statements 41, 43, 45-48, 51, and 53-56 (Instagram’s harm to young users) are false or misleading. The Court therefore turns to whether Plaintiffs have alleged scienter. Defendants seek dismissal on the ground that Plaintiffs do not plead particularized facts supporting a strong inference of scienter. Mot. at 25-27, 34-35, 40-41, 45. To establish scienter, the complaint must “state with particularity facts giving rise to a strong inference that the defendant acted with the required state of mind.” 15 U.S.C. § 78u– 4(b)(2)(A). The required state of mind is “a mental state that not only covers ‘intent to deceive, manipulate, or defraud,’ but also ‘deliberate recklessness.’ ” Schueneman v. Arena Pharms., Inc., 840 F.3d 698, 705 (9th Cir. 2016) (internal citations omitted). Deliberate recklessness is “ ‘an extreme departure from the standards of ordinary care,’ which ‘presents a danger of misleading buyers or sellers that is either known to the defendant or is so obvious that the actor must have been aware of it.’ ” In re Alphabet, Inc. Sec. Litig., 1 F.4th 687, 701 (9th Cir. 2021) (emphasis in original) (quoting Nguyen v. Endologix, Inc., 962 F.3d 405, 414 (9th Cir. 2020)). The “strong inference” required by the PSLRA “must be more than merely ‘reasonable’ or ‘permissible’—it must be cogent and compelling, thus strong in light of other explanations.” Tellabs, 551 U.S. at 324. “Facts showing mere recklessness or a motive to commit fraud and opportunity to do so provide some reasonable inference of intent, but are not sufficient to establish a strong inference of deliberate recklessness.” In re VeriFone Holdings, 704 F.3d at 701. “A court must compare the malicious and innocent inferences cognizable from the facts pled in the complaint, and only allow the complaint to survive a motion to dismiss if the malicious inference is at least as compelling as any opposing innocent inference.” Zucco Partners, 552 F.3d at 991 (9th Cir. 2009); see Nguyen, 962 F.3d at 415. In evaluating whether a complaint satisfies the “strong inference” requirement, courts must consider the allegations and other relevant material “holistically,” not “scrutinized in isolation.” In re VeriFone Holdings, 704 F.3d at 701-02 (citing Tellabs, 551 U.S. at 323, 326). Because scienter is a subjective inquiry, “the ultimate question is whether the defendant knew his or her statements were false, or was 2010). The Court discusses Plaintiffs’ scienter allegations with respect to the two remaining categories of statements: cross-check and Instagram’s harm to young users. 1. Scienter Allegations as to Cross-Check Plaintiffs’ scienter allegations about cross-check fall into three general categories: (a) Defendants’ personal involvement in the subject matter and Defendants’ contemporaneous knowledge; (b) the core operations doctrine; (c) motive to conceal the truth. The Court discusses each below. a. Personal Involvement and Contemporaneous Knowledge To plead scienter, a complaint may plead a “combination” of facts, none of which need to be from confidential witnesses, internal reports, or other specific sources. See In re Alphabet, 1 F.4th at 707 (collecting authority for proposition “[a]llegations of suspicious stock sales or information from confidential witnesses are not needed”). However, the PSLRA demands “particular allegations which strongly imply Defendants’ contemporaneous knowledge that the statement was false when made.” Berson v. Applied Signal Tech., Inc., 527 F.3d 982, 989 (9th Cir. 2008) (emphasis in original) (citation omitted). Here, Plaintiffs allege that Defendants’ personal involvement in cross-check supports an inference of scienter because Defendants directly participated in the enforcement of the rules and touted their personal involvement in cross-check. Response at 22. Plaintiffs allege that Zuckerberg admitted to being personally involved in the decision to remove former President Trump’s content, that he generally seeks out content himself, and that Zuckerberg said Clegg was also involved in Meta’s cross-check decision-making process. FAC ¶¶ 101, 111-12, 545-56. Clegg admitted to overseeing company policies about hate speech rules work. FAC ¶ 546. Plaintiffs also allege that a Meta report revealed that Meta’s “public policy team” was involved in the enforcement process, and that Bickert and Stone are on the public policy team. FAC ¶¶ 58, 62, 111. An individual’s involvement in the subject matter of the misstatements may support an inference of scienter. See City of Miami Gen. Employees’ & Sanitation Employees’ Ret. Tr. v. RH, in the product lunch, made last-minute changes to designs, received weekly reports showing that inventory was out of stock, and “had the final word” on product ordering). Plaintiffs’ allegations that Zuckerberg had hands-on involvement in the decision-making process for cross-check raises an inference of scienter, as do the allegations that Zuckerberg named Clegg as someone involved in the decision-making and Clegg’s admission that he oversaw company policies in this area. See id. However, as to Bickert and Stone, Plaintiffs fail to point to any factual allegations supporting a strong inference of scienter. Plaintiffs argue that Bickert and Stone were personally involved in the cross-check process based on allegations that Bickert posted the Community Standards and stated the importance of “how we decide” what goes on Facebook and that “we remove content” that violates standards. Response at 23 (citing FAC ¶¶ 88, 90) (emphasis in original). Stone stated that cross-check was merely a second layer of review and that there were not two systems of justice. Id. (citing FAC ¶ 289). However, without allegations that Bickert and Stone knew that these statements were false or misleading, or had access to or reviewed the reports showing contradictory information, these allegations cannot support a strong inference of scienter as to them. See S. Ferry LP, No. 2 v. Killinger, 542 F.3d 776, 784 (9th Cir. 2008) (alleging management’s important role without detailed allegations of “actual exposure to information” is insufficient to show scienter); see In re Eargo, Inc. Sec. Litig., 656 F. Supp. 3d 928, 948 (N.D. Cal. 2023) (“vague” allegations about “internal review” insufficient to establish scienter where plaintiff provided no details about what review showed). b. Core Operations Plaintiffs also allege that content enforcement was a “core product” that was widely studied and communicated through Meta. FAC ¶¶ 518-22. They allege that a Meta audit found that whitelisting was “pervasive, touching almost every area of the company,” 45 teams were involved in whitelisting, and an internal review found that Meta is “not actually doing what we say we do publicly.” FAC ¶¶ 104, 107-08, 114. These general allegations of an internal review or audit are insufficient as Plaintiffs provide no detail about who drafted the documents, who Litig., 452 F. Supp. 2d 1005, 1024-25 (S.D. Cal. 2006) (allegations that Defendants were informed of problems through weekly reports were insufficient where plaintiffs failed to allege “when these reports were generated and distributed,” who received them, and the contents of the reports). Indeed, Plaintiffs’ selective quotations of conclusions from reports published in the Wall Street Journal do not indicate who drafted the reports, what supported the conclusions quoted by Plaintiffs, or whether any of the Defendants reviewed the reports. These allegations do not support a strong inference of scienter. As to Plaintiffs’ core operations theory, Plaintiffs’ vague argument – without support in the FAC – that Meta’s content enforcement practices “concerned its core product and were closely monitored,” Response at 23, also cannot support a finding of scienter. c. Assertions of Improper Motive Plaintiffs also argue that the “desire to maximize Meta’s profitability” motivated Defendants to make false statements because their conduct drove up user engagement. FAC ¶ 573. “[G]eneralized assertions of motive, without more, are inadequate to meet the [PSLRA’s] heightened pleading requirements.” Lipton, 284 F.3d at 1038 (finding that the “desires to obtain favorable financing and to expand abroad are in themselves ordinary and appropriate corporate objectives” and without more cannot support a finding of fraud). Accordingly, Defendants’ broad financial motives cannot support a finding of scienter. “[I]f no individual allegation is sufficient, [courts] conduct a ‘holistic’ review of the same allegations to determine whether the insufficient allegations combine to create a strong inference of intentional conduct or deliberate recklessness.” City of Dearborn Heights, 856 F.3d at 620. The FAC alleges that Defendants Zuckerberg and Clegg were individually involved in the cross- check decision-making process, FAC ¶¶ 101, 111-12, 545-46, and made detailed statements about cross-check, which is sufficient to draw a strong inference that they acted with scienter. See City of Miami, 302 F. Supp. 3d at 1044; see also Reese v. Malone, 747 F.3d 557, 576 (9th Cir. 2014), overruled on other grounds by City of Dearborn Heights Act 345 Police & Fire Ret. Sys. v. Align Tech., Inc., 856 F.3d 605 (9th Cir. 2017) (finding strong inference of scienter where defendant’s 2. Scienter Allegations as to Instagram’s Harm to Young Users As to Instagram’s harm to young users, Plaintiffs allege scienter based on: (a) Defendants’ awareness of contemporaneous reports; (b) the motive to conceal the research; and (c) the prominence of these issues and Defendants speaking about them. The Court considers each below. a. Contemporaneous Knowledge Plaintiffs argue that Defendants were aware of reports or data about Instagram harming teen mental health because the research was available to all Meta employees. FAC ¶ 534. However, simply pleading that Defendants had “access to internal data” or reports cannot support an inference that Defendants were aware of the research. See Lipton, 284 F.3d at 1026 (finding no inference of scienter where plaintiffs “assert[ed] in conclusory terms that defendants had access to internal data demonstrating a decline in sales,” but failed to identify any internal reports of such data or plead “the contents of any such report or the purported data”). Plaintiffs also allege that top executives reviewed “deep dive” research about Meta’s social media platforms harming teenagers’ mental health, and that a 2020 presentation to Zuckerberg cited this information. FAC ¶ 534. Again, Plaintiffs fail to allege the contents of the research that top executives reviewed (or which top executives reviewed the research). See City of Dearborn Heights, 856 F.3d at 620 (allegations of scienter were not sufficiently particular where plaintiff alleged that defendants had “direct access to the data room” but did not allege that defendants personally accessed the data room or that the information was disclosed to them); see also Sanders v. Realreal, Inc., 2021 WL 1222625, at *15 (N.D. Cal. Mar. 31, 2021) (no scienter where some upper-level employees had “clear access” to the relevant spreadsheets). Accordingly, Plaintiffs’ allegations of contemporaneous knowledge fail to support a strong inference of scienter. b. Motives to Conceal Instagram Research Plaintiffs also allege that Defendants were motivated to hide Meta’s internal research because they knew that disclosing it would reduce the number of young users on their platforms and harm their profits. FAC ¶ 581. As with the scienter allegations for cross-check, above, scienter. See Lipton, 284 F.3d at 1038. c. Prominence of Issues and Speaking About Them Plaintiffs also allege that the Court can infer scienter because Defendants publicly acknowledged their awareness of Meta’s research in this area and purported to speak knowledgeably about it. Response at 41 (citing FAC ¶¶ 244, 246, 251-52, 488, 492, 495-96, 545- 55). They add that the issues were of significant prominence given that Meta commissioned multi-year research and Congress raised these issues with Meta at hearings on Instagram’s harm to young users. Id. (citing FAC ¶¶ 241-43, 255-57, 571-72, 580). Zuckerberg testified to Congress that Meta is studying its platform’s harm to children and testified about research he had seen. FAC ¶¶ 241, 554-55. Newton, Instagram’s head of public policy, stated that the Wall Street Journal article improperly focused on limited findings and cast them in a negative light. FAC ¶ 492. Davis, Meta’s Head of Safety, testified before a Senate Committee titled “Protecting Kids Online: Facebook, Instagram, and Mental Health Harms” about Meta’s internal research in this area. FAC ¶ 495. Where Defendants directly address a specific issue or data, particularly in relation to a government inquiry, courts may infer that Defendants were aware of the issue or data referenced. See Reese, 747 F.3d at 571 (inferring that Defendant had access to the data given Defendant’s position and the fact that she made a statement specifically addressing the data at issue in the context of a public and government inquiry). Like in Reese, Defendants Zuckerberg, Newton, and Davis publicly and directly addressed the research on Instagram’s harm to young users, suggesting that they had access to the relevant information. See id.; see also S. Ferry, 542 F.3d at 785 (holding that “detailed and specific allegations about management’s exposure to factual information within the company” support an inference of scienter). Indeed, “an assertion that Defendants were unaware of an alleged issue can be directly contradicted by the fact that [they] specifically addressed it in [their] statement[s].” Shenwick, 282 F. Supp. at 1147. However, the allegations as to Mosseri (Instagram’s CEO) and Diwanji (Vice President of Youth Products) do not support a finding of scienter. In questions to Mosseri about whether social 251. As to Diwanji, the only allegation is that in a blog post, she claimed that Meta had convened experts to develop a comprehensive plan on child safety and well-being. FAC ¶ 488. Those allegations fail to support a finding that either Mosseri or Diwanji was aware of Meta’s internal research. See, e.g., Glazer Capital Mgmt. v. Magistri, 549 F.3d 736, 743-49 (9th Cir. 2008) (no strong inference of scienter on the part of the company CEO in the absence of facts showing he was personally aware of illegal payments or that he was actively involved in details of Asian sales); see also In re Enovix Corp. Sec. Litig., 2024 WL 349269, at *15 (N.D. Cal. Jan. 30, 2024) (concluding that “the consolidated complaint does not state with specificity what defendants knew and when”) (emphasis in original). Given that Zuckerberg, Newton, and Davis specifically addressed Meta’s research about Instagram harming children, the Court finds Plaintiffs have alleged a strong inference of scienter. Indeed, Plaintiffs must “raise a strong inference of fraud . . . [but] do not have to conclusively eliminate all doubt.” SEB Inv. Mgmt. AB v. Wells Fargo & Co., 2024 WL 3579322, at *11 (N.D. Cal. July 29, 2024) (quoting In re LDK Solar Sec. Lit., 584 F. Supp. 2d 1230, 1255-56 (N.D. Cal. 2008)).11 C. Loss Causation Under the PSLRA, a plaintiff must show “loss causation.” That is, they must show a “causal connection between the deceptive acts that form the basis for the claim of securities fraud and the injury suffered by the plaintiff.” In re Gilead, 536 F.3d at 1055 (internal quotation marks and citation omitted). The misrepresentation need not be the “sole” cause of the loss, but it must be a “substantial” one. See id. Typically, loss causation is shown through “corrective disclosures” that reveal the truth to the market and “cause[ ] the company’s stock price to drop and investors to 11 Defendants allege that Plaintiffs have failed to allege scienter as to the Individual Defendants. For the reasons stated above, the Court cannot agree as to Zuckerberg, Clegg, Newton, and Davis. See Supra II(B). However, Plaintiffs fail to plead with particularity that the remaining Individual Defendants – Mosseri, Bickert, Stone, Wehner, LeCun, Rosen, or Diwanji – made a false or misleading statement with scienter. See Veal v. Lendingclub Corp., 2020 WL 3128909, at *15 (N.D. Cal. June 12, 2020), aff’d, 2021 WL 4281301 (9th Cir. Sept. 21, 2021) (dismissing claims where the complaint “lumps all Defendants together with respect to the scienter allegations”). lose money.” Lloyd, 811 F.3d at 1209. The test for loss causation “requires no more than the familiar test for proximate cause.” Mineworkers’ Pension Scheme v. First Solar Inc., 881 F.3d 750, 753 (9th Cir. 2018) (citation omitted). Here, Plaintiffs have adequately alleged loss causation. As to the cross-check statements, Plaintiffs allege that disclosures on September 13, 21, and 28 and October 21, 2021, revealed information about cross-check that proximately caused Meta’s stock to decline. FAC ¶¶ 279-90, 322-27, 333-39, 514. Defendants argue that they had previously made public statements about cross-check, so the disclosures were not “new.” Mot. at 28 (citing FAC ¶¶ 90, 105-06). However, the FAC alleges that those earlier “disclosures” were false or misleading. See FAC ¶¶ 90, 105-06. Thus, they did not disseminate knowledge to the market. Moreover, proof that “news of the truth credibly entered the market and dissipated the effects of prior misstatements” is an issue for trial or summary judgment. Amgen Inc. v. Connecticut Ret. Plans & Tr. Funds, 568 U.S. 455, 482 (2013) (citation and internal brackets omitted). As to Instagram harming young users, Plaintiffs allege that disclosures on September 14 and 28 caused Meta’s stock price to fall nearly $13 a share. FAC ¶¶ 291-301, 329-332, 334. Further, on October 3, 2021, the Meta whistleblower identified herself as the credible source of previously-disclosed information, and revealed additional information about Facebook’s research findings, causing Meta’s stock to drop $16.78 per share. FAC ¶ 514. “[A] stock price drop [that] comes immediately after the revelation of fraud can help to rule out alternative causes.” Mineworkers’ Pension Scheme, 881 F.3d at 754. That is so here. Plaintiffs have thus plausibly alleged that the disclosures about the broad application of the cross-check policy and Instagram’s research on its harm to young users were at least a substantial cause of the loss of value. See In re QuantumScape, 580 F. Supp. 3d at 742. III. SECTION 20(A) CLAIM Plaintiffs also bring a Section 20(a) claim against Defendants Zuckerberg, Wehner, Clegg, Mosseri, Rosen, Stone, Davis, Newton, LeCun, Bickert, and Diwanji. FAC ¶¶ 609-616. Under Section 20(a) of the Exchange Act, “certain ‘controlling’ individuals [are] also liable for violations of [S]ection 10(b) and its underlying regulations.” Zucco Partners, 552 F.3d at 990 (citing 15 1 corporation who has violated the securities laws will be jointly and severally liable to the plaintiff, 2 as long as the plaintiff demonstrates ‘a primary violation of federal securities law’ and that ‘the 3 defendant exercised actual power or control over the primary violator.’ ” Jd. (citation omitted). 4 Plaintiffs’ allegations about Individual Defendants’ “title and responsibilities” are “sufficient to 5 establish control at the motion to dismiss stage.” Kyung Cho v. UCBH Holdings, Inc., 890 F. 6 Supp. 2d 1190, 1205 (N.D. Cal. 2012) (citation omitted) (citing cases). Because Lead Plaintiff's 7 Section 20(a) claim rises and falls with its claim under Section 10(b), the claim is DISMISSED 8 WITH LEAVE TO AMEND to the same extent as the Section 10(b) claim. The Section 20(a) 9 claim otherwise survives consistent with the Court’s analysis of the viable portion of the Section 10 10(b) claim. See Zaidi v. Adamas Pharms., Inc., 650 F. Supp. 3d 848, 865 (N.D. Cal. 2023). 12 Accordingly, the Court GRANTS in part and DENIES in part the motion to dismiss. The 13 Court DISMISSES the claims based on Statements 19, 22, and 33 as a matter of law. The Court 14 DENIES the motion to dismiss the claims based on Statements 8 and 11 as to Defendants Meta, 3 15 Zuckerberg, and Clegg, and Statements 41, 43, 45-48, and 51-56 as to Defendants Meta, a 16 Zuckerberg, Newton, and Davis. The Court DISMISSES with leave to amend the claims based 3 17 on the remaining statements and as to the remaining Defendants. If Plaintiffs elect to amend, any 18 amended complaint must be filed by October 30, 2024. No additional parties or claims may be 19 added without leave of Court or stipulation of Defendants. 21 Dated: September 30, 2024 22 Wed / 7

ARACELI MARTINEZ-OLGUIN 24 United States District Judge 25 26 27 28